The Advanced Scientific Computing Research (ASCR) program in the US Department of Energy (DOE) Office of Science is organizing a Workshop on Energy-Efficient Computing for Science (EECS).
Key dates:
- July 22, 2024: Deadline for position paper submission
- August 2, 2024: Notification of position acceptance
- September 9-12, 2024: Workshop (greater Washington, DC area)
- Workshop Website: https://www.orau.gov/
2024EECWorkshop
Energy efficiency involves coordination across all the interoperating components of a computing system—in particular, applications, algorithms, system software, programming models, data management, and the hardware on which they run. Looking 10-15 years into the future, the goal is to dramatically lower the energy costs of the computational platforms (from the data center to the edge) serving DOE science while expanding the capabilities of these systems, broadening their applicability to science challenges of interest to DOE and the nation.
For this call for papers, any metric relevant to energy efficiency is in scope but nominally energy efficiency is defined as performance throughput per unit power, considering not only the power consumed by compute resources but also networking, data storage and movement, and cooling. The EECS workshop will explore how the energy efficiency of scientific applications of interest to DOE can be enhanced through advances in the aforementioned technology areas, but also through close interactions among hardware designers, software-stack developers, and application teams in which all components adapt to maximize energy efficiency (e.g., via co-design).
The workshop will include breakout sessions, with every attendee expected to participate actively in the discussions. Afterward, workshop attendees — from DOE, industry, and academia — will produce a report for ASCR summarizing the findings made during the workshop and identifying priority research directions.
Invitation
We invite community input through two-page position papers identifying and discussing key challenges and opportunities in improving energy efficiency for scientific computing and promising research avenues to meet these challenges. In addition to providing an avenue for identifying workshop participants, these position papers will be used to shape the workshop agenda, identify panelists, and contribute to the workshop report. Position papers should not describe the authors’ current or planned research, contain material that should not be disclosed to the public, nor should they recommend specific solutions or discuss narrowly focused research topics. Instead, they should aim to improve the community’s shared understanding of the problem space, identify challenging research directions, and help to stimulate discussion.
Authors are not required to have a history of funding by ASCR. One author of each selected submission will be invited to participate in the workshop.
By submitting a position paper, authors consent to publishing it publicly.
Submission Guidelines
Position Paper Structure and Format
Position papers should follow the following format:
- Title
- Authors (with affiliations and email addresses)
- Topic: one or more of the following: applications, algorithms, modeling and simulation, system software, programming systems, data management, architectures, emerging technologies
- Challenge(s): Identify limitations of state-of-the-art practice with examples
- Opportunity: Describe how the identified challenges may be addressed, whether it is through new tools and techniques, new technologies, or new groups collaborating in an integrated manner
- Timeliness or maturity: Why now? What breakthrough or change makes progress possible now where it wasn’t possible before? What will be the impact of success?
- References
Each position paper must be no more than two pages, including figures and references, using 1” margins and an 11-point font. The paper may include any number of authors, but contact information for a single author who can present the position paper at the workshop must be provided with the submission. There is no limit to the number of position papers that an individual or group can submit. Authors are strongly encouraged to follow the previously outlined structure. Papers should be submitted in PDF format using the designated page on the workshop website.
Areas of Emphasis
We are seeking submissions along the following general directions:
- Approaches to development of energy-efficient applications
- Hardware technologies (e.g., computing, networking, data storage, cooling) and their impact on energy efficiency.
- Energy-efficient algorithms for scientific computing at scale.
- Energy-efficient management of exabyte-scale data storage, including streaming data.
- Novel computing paradigms (hardware, architectures, algorithms) that prioritize energy efficiency.
- Software strategies, programming methods, and tools that improve or measure energy efficiency.
- Benchmarks and metrics for measurements of energy-efficient performance for applications, algorithms and architectures.
- Exploring how energy efficiency can emerge from a perspective considering hardware, software, and algorithms together.
- Insights into energy efficiency for workflows arising from scientific experiments, on supercomputers, executing on edge systems or to support large-scale scientific instruments.
- Considerations for energy efficiency in different deployment modalities: cloud, on-premises, edge, etc.
Notional Questions
Position papers should present a view on how one or more aspects of hardware, the software stack, data management, algorithms, or scientific applications can enable substantial improvements in energy efficiency, perhaps taking inspiration from some of the following:
- What emerging hardware technologies, software technologies, or algorithmic advancements will facilitate rapid gains in the energy efficiency of scientific applications and in what time frame? What are the tools and techniques needed to enable the successful integration of these advancements into production systems? What are the risks associated with each technology?
- How to balance breadth of supported applications versus energy savings for particular applications (for instance with specialization)? Does this vary by system type?
- As scientific workflows and applications become more heterogeneous (e.g., graph analytics, optimization, streaming data analysis), what new challenges and opportunities does this present for energy efficiency?
- With artificial intelligence and machine learning becoming widely used in scientific workflows and applications, what new challenges and opportunities does this present for energy efficiency?
- What is the potential for new accelerator technologies and chiplets to reduce energy requirements, and what steps are needed to take maximum advantage of them?
- How can we further the state of the art in efficient and flexible open-source hardware, modeling, and simulation tools that can underpin new approaches to energy efficiency?
- How can algorithm design factor in energy efficiency instead of / alongside speed?
- How do we revisit benchmarks to properly factor in energy as a consideration of systems and applications?
- When should overall energy be considered, and when should peak performance be the priority? How should we account for cooling costs in the energy consumption?
- What changes are needed in managing data to facilitate energy savings at the system or facility level?
- How does an emphasis on energy savings change how large-scale computing resources are managed? Can this impact individual application performance?
- What changes in scientific applications and underlying numerical algorithms are needed to exploit emerging hardware and system software capabilities?
- What should application, system software developers, and hardware designers do differently when prioritizing energy efficiency?
- How could the co-design of scientific applications and hardware/software platforms accelerate the adoption of or raise the effectiveness of energy-efficient computing solutions?
- What role does system software play in energy efficiency? What new roles must it take on?
- Open-source software is supported by long-standing DOE policy. Can open-source software together with open-source hardware support energy efficiency for science?
- What is the role of DOE in energy efficiency for the computing continuum as a whole?
- What programming systems concepts facilitate energy efficiency?
Selection
Submissions will be reviewed by the workshop’s organizing committee using criteria of overall quality, relevance, likelihood of stimulating constructive discussion, and ability to contribute to an informative workshop report. Unique positions that are well presented and emphasize potentially transformative research directions will be given preference.
Workshop Organizers
Co-Chairs
- Rob Ross, Argonne National Laboratory
- Jeffrey Vetter, Oak Ridge National Laboratory
Organizing Committee
- Brad Aimone, Sandia National Laboratories
- George Amvrosiadis, Carnegie Mellon University
- Jim Ang, Pacific Northwest National Laboratory
- Hyesoon Kim, Georgia Institute of Technology
- Michel Kinsy, Arizona State University
- George Michelogiannakis, Lawrence Berkeley National Laboratory
For meeting technical questions, please contact: Todd Munson, Todd.Munson@science.doe.gov
